NBA Analytics Blog · Aug 21, 2026 · 7 min read
An NBA statistics guide for analysis: per-possession ratings, the four factors, pace, rest and availability — which numbers predict and which only describe.
By Sportslyx · Published
The NBA publishes more statistics than any other league, and most of them describe rather than predict. This guide sorts the useful ones from the decorative ones: how to put everything on a per-possession basis, which measures explain the spread, why rest and availability dominate the late line, and what player numbers say about a game when a starter sits. Each section notes where the statistic appears on a Sportslyx page and how the read is graded.
A team that scores 118 points a night is not necessarily a good offense; it may simply play fast. Pace — possessions per 48 minutes — varies enough between teams that raw scoring and raw defense are unreadable without it. The fix is to express everything per 100 possessions. Offensive rating is points scored per 100 possessions, defensive rating is points allowed per 100, and net rating is the difference. Net rating adjusted for schedule strength is the single best descriptor of team quality in the league and correlates with the spread far better than record.
Every possession ends in one of a small number of ways, and the four factors describe how a team wins them: effective field-goal percentage (shooting, with threes weighted), turnover rate, offensive rebounding rate and free-throw rate. Shooting matters most, turnovers next. Reading the four factors for both teams — a team’s own and what it allows — tells you where an efficiency edge comes from, which is more useful than knowing that one exists.
| Statistic | Descriptive or predictive? | Note |
|---|---|---|
| Points per game | Descriptive | Contains pace; use ratings instead |
| Net rating (per 100) | Predictive | Best single team-quality measure |
| Effective FG% | Predictive | Largest of the four factors |
| Turnover rate | Predictive | Second-largest factor |
| Win–loss record | Descriptive | Contains close-game luck and past injuries |
| Three-point percentage (recent) | Mostly noise | Regresses quickly; volume persists, accuracy less so |
| Rest days / back-to-back | Predictive | Small but consistent effect, larger for older rosters |
Pace tells you about the shape of the game, not who wins it. Two fast teams produce a high total; a fast team against a slow one usually lands in between, tilted toward whoever controls tempo. Pace is the first input for a total and a secondary input for a side — it multiplies an efficiency edge rather than creating one.
Basketball is played almost every night, and fatigue is measurable. Teams on the second night of a back-to-back, especially on the road, underperform their ratings by a small but consistent margin, and the effect grows with roster age and with the travel involved. The line usually accounts for it; the question is whether it has accounted for the specific roster. The Sportslyx package records rest days for both teams alongside the line.
No sport confirms its lineups later than the NBA, and no late news moves a line more. A single star’s absence can shift a spread by several points; two can flip the favorite. Load management means healthy players sit on back-to-backs, and the injury report is only reliable close to tip-off. The Sportslyx breakdown is pre-generated and cached, so a morning version may have been written for a different lineup; the win probability on the card follows the current line, and the data tier says how much of the report the package had when the text was written.
Availability also changes who benefits. When a high-usage player sits, minutes and shots go somewhere, and the statistics that predict where — usage rate, minutes per game with and without the star — are the ones to read when a starter sits.
Home-court advantage in the NBA has shrunk over the past decade but is still worth roughly two to three points. It is not uniform: altitude venues and long road trips add to it. Separately, garbage time distorts per-game numbers; a 25-point win that ends 12 after bench minutes looks closer than it was. Net rating filtered for competitive minutes is a cleaner measure when it is available.
The three-point shot is the largest source of game-to-game noise in the modern NBA. A team shooting 44% from three over its last five games is not a 44% team; three-point accuracy regresses toward a player’s and a team’s longer-run rate quickly, while three-point volume is stable. When a recent run of results looks better than the ratings, check the three-point percentage first.
Player numbers matter to a game read through the same logic as team numbers: minutes, usage, matchup and pace. A scorer’s output is a function of how many possessions he is on the floor for, how often the offense runs through him, the opponent’s defensive efficiency at his position and the expected pace of the game. When he sits, those possessions and shots go to someone, and the team’s rating with and without him is the number that says how much the side and the total should move.
One product note belongs here. Player props are covered for MLB only on Sportslyx; an NBA read is on the game — moneyline, spread or total — so the player statistics in this guide are inputs to that read, not a separate market. A read that depends on a game-time decision is a read that may describe a lineup that never takes the floor, which is why the report closest to tip-off is the one to trust.
A moneyline read on Sportslyx settles on the final score including overtime; a postponed game is void; player props are covered for MLB only, so there is no prop leg in the NBA record. Results aggregate over one calendar month in UTC, with hit rate excluding pushes and ROI at a flat stake, on the pick tracker. The NBA’s long season gives the steadiest monthly record on the site — and empty summers, which are the off-season, not a missing file. The statistics discussed here live on the NBA statistics hub; the NBA predictions board shows tonight’s probabilities.
Net rating per 100 possessions, adjusted for schedule strength. It removes pace, captures both ends of the floor and correlates with the spread far better than win–loss record.
A small but consistent amount — a point or two against a team’s ratings — larger on the road and for older rosters. The market usually prices it; the question is whether the specific lineup and travel have been accounted for.
Not at present. Player props are covered for MLB only, from the public feed. An NBA read is on the game — moneyline, spread or total — and the player numbers in this guide matter as inputs to that read, above all for where the minutes and shots go when a starter sits.
Take this guide to the NBA board: the 7-day Pro trial opens the AI breakdowns, the market-derived probabilities, the daily company board, Combo Builder, Studio and the pick tracker.
7-day trial · $0 today · card required · cancel anytime before the trial ends
Sportslyx produces statistical estimates and analytical breakdowns for information and entertainment. Nothing on this page is betting advice, and no outcome is certain. 18+ only. Please play responsibly — see our Responsible Play page.
